Follow-Me Project¶

Congratulations on reaching the final project of the Robotics Nanodegree!

Previously, you worked on the Semantic Segmentation lab where you built a deep learning network that locates a particular human target within an image. For this project, you will utilize what you implemented and learned from that lab and extend it to train a deep learning model that will allow a simulated quadcopter to follow around the person that it detects!

Most of the code below is similar to the lab with some minor modifications. You can start with your existing solution, and modify and improve upon it to train the best possible model for this task.

You can click on any of the following to quickly jump to that part of this notebook:

Data Collection
FCN Layers
Build the Model
Training
Prediction
Evaluation

Data Collection ¶

We have provided you with a starting dataset for this project. Download instructions can be found in the README for this project's repo. Alternatively, you can collect additional data of your own to improve your model. Check out the "Collecting Data" section in the Project Lesson in the Classroom for more details!

import os
import glob
import sys
import tensorflow as tf

from scipy import misc
import numpy as np

from tensorflow.contrib.keras.python import keras
from tensorflow.contrib.keras.python.keras import layers, models

from tensorflow import image

from utils import scoring_utils
from utils.separable_conv2d import SeparableConv2DKeras, BilinearUpSampling2D
from utils import data_iterator
from utils import plotting_tools 
from utils import model_tools

FCN Layers ¶

In the Classroom, we discussed the different layers that constitute a fully convolutional network (FCN). The following code will introduce you to the functions that you need to build your semantic segmentation model.

Separable Convolutions¶

The Encoder for your FCN will essentially require separable convolution layers, due to their advantages as explained in the classroom. The 1x1 convolution layer in the FCN, however, is a regular convolution. Implementations for both are provided below for your use. Each includes batch normalization with the ReLU activation function applied to the layers.

def separable_conv2d_batchnorm(input_layer, filters, strides=1):
    output_layer = SeparableConv2DKeras(filters=filters,kernel_size=3, strides=strides,
                             padding='same', activation='relu')(input_layer)
    
    output_layer = layers.BatchNormalization()(output_layer) 
    return output_layer

def conv2d_batchnorm(input_layer, filters, kernel_size=3, strides=1):
    output_layer = layers.Conv2D(filters=filters, kernel_size=kernel_size, strides=strides, 
                      padding='same', activation='relu')(input_layer)
    
    output_layer = layers.BatchNormalization()(output_layer) 
    return output_layer

Bilinear Upsampling¶

The following helper function implements the bilinear upsampling layer. Upsampling by a factor of 2 is generally recommended, but you can try out different factors as well. Upsampling is used in the decoder block of the FCN.

def bilinear_upsample(input_layer):
    output_layer = BilinearUpSampling2D((2,2))(input_layer)
    return output_layer

Build the Model ¶

In the following cells, you will build an FCN to train a model to detect and locate the hero target within an image. The steps are:

Create an encoder_block
Create a decoder_block
Build the FCN consisting of encoder block(s), a 1x1 convolution, and decoder block(s). This step requires experimentation with different numbers of layers and filter sizes to build your model.

Encoder Block¶

Create an encoder block that includes a separable convolution layer using the separable_conv2d_batchnorm() function. The filters parameter defines the size or depth of the output layer. For example, 32 or 64.

def encoder_block(input_layer, filters, strides):
    
    # TODO Create a separable convolution layer using the separable_conv2d_batchnorm() function.
    sep_conv1=separable_conv2d_batchnorm(input_layer,filters,strides)
    output_layer=sep_conv1
    return output_layer

Decoder Block¶

The decoder block is comprised of three parts:

A bilinear upsampling layer using the upsample_bilinear() function. The current recommended factor for upsampling is set to 2.
A layer concatenation step. This step is similar to skip connections. You will concatenate the upsampled small_ip_layer and the large_ip_layer.
Some (one or two) additional separable convolution layers to extract some more spatial information from prior layers.

def decoder_block(small_ip_layer, large_ip_layer, filters):
    
     # TODO Upsample the small input layer using the bilinear_upsample() function.
    up_layer1=bilinear_upsample(small_ip_layer)
    
    # TODO Concatenate the upsampled and large input layers using layers.concatenate
    up_layer1=layers.concatenate([up_layer1,large_ip_layer])
    # TODO Add some number of separable convolution layers
    up_layer2=separable_conv2d_batchnorm(up_layer1, filters, strides=1)
#     up_layer3=separable_conv2d_batchnorm(up_layer2, filters, strides=1)
    
    output_layer=up_layer2
    
    return output_layer

Model¶

Now that you have the encoder and decoder blocks ready, go ahead and build your FCN architecture!

There are three steps:

Add encoder blocks to build the encoder layers. This is similar to how you added regular convolutional layers in your CNN lab.
Add a 1x1 Convolution layer using the conv2d_batchnorm() function. Remember that 1x1 Convolutions require a kernel and stride of 1.
Add decoder blocks for the decoder layers.

def fcn_model(inputs, num_classes):
    
    # TODO Add Encoder Blocks. 
    # Remember that with each encoder layer, the depth of your model (the number of filters) increases.
    print("Feature layer", inputs.shape)
    encoder_1_layer = encoder_block(inputs, 32, 2)
    encoder_2_layer = encoder_block(encoder_1_layer, 64, 2)
    encoder_3_layer = encoder_block(encoder_2_layer, 128, 2)
    print("Encoder 1", encoder_1_layer.shape)
    print("Encoder 2", encoder_2_layer.shape)
    print("Encoder 3", encoder_3_layer.shape)
    
    # TODO Add 1x1 Convolution layer using conv2d_batchnorm().
    conv_1_1_layer = conv2d_batchnorm(encoder_3_layer, 128, 1, 1)
    print("1x1 Conv", conv_1_1_layer.shape)
    
    # TODO: Add the same number of Decoder Blocks as the number of Encoder Blocks
    decoder_1_layer = decoder_block(conv_1_1_layer, encoder_2_layer, 128)
    print("Decoder 1", decoder_1_layer.shape)
    decoder_2_layer = decoder_block(decoder_1_layer, encoder_1_layer, 64)
    print("Decoder 2", decoder_2_layer.shape)
    decoder_3_layer = decoder_block(decoder_2_layer, inputs, 32)
    print("Decoder 3", decoder_3_layer.shape)
    
    x=decoder_3_layer
    
    output_layer=layers.Conv2D(num_classes, 1, activation='softmax', padding='same')(x)
    print("Output_layer", output_layer.shape)
    
    
    # The function returns the output layer of your model. "x" is the final layer obtained from the last decoder_block()
    return output_layer

Training ¶

The following cells will use the FCN you created and define an ouput layer based on the size of the processed image and the number of classes recognized. You will define the hyperparameters to compile and train your model.

Please Note: For this project, the helper code in data_iterator.py will resize the copter images to 160x160x3 to speed up training.

def score_all():
    run_num = 'run_1'

    val_with_targ, pred_with_targ = model_tools.write_predictions_grade_set(model,
                                        run_num,'patrol_with_targ', 'sample_evaluation_data') 
    val_no_targ, pred_no_targ = model_tools.write_predictions_grade_set(model, 
                                        run_num,'patrol_non_targ', 'sample_evaluation_data') 
    val_following, pred_following = model_tools.write_predictions_grade_set(model,
                                        run_num,'following_images', 'sample_evaluation_data')
    true_pos1, false_pos1, false_neg1, iou1 = scoring_utils.score_run_iou(val_following, pred_following)
    true_pos2, false_pos2, false_neg2, iou2 = scoring_utils.score_run_iou(val_no_targ, pred_no_targ)
    true_pos3, false_pos3, false_neg3, iou3 = scoring_utils.score_run_iou(val_with_targ, pred_with_targ)
    true_pos = true_pos1 + true_pos2 + true_pos3
    false_pos = false_pos1 + false_pos2 + false_pos3
    false_neg = false_neg1 + false_neg2 + false_neg3

    weight = true_pos/(true_pos+false_neg+false_pos)
    print(weight)
    final_IoU = (iou1 + iou3)/2
    print(final_IoU)
    final_score = final_IoU * weight
    print(final_score)

# Run One
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""

image_hw = 160
image_shape = (image_hw, image_hw, 3)
inputs = layers.Input(image_shape)
num_classes = 3

# Call fcn_model()
output_layer = fcn_model(inputs, num_classes)

learning_rate = 0.01
batch_size = 32
num_epochs = 40
steps_per_epoch = int(4131/batch_size)-1#200
validation_steps = int(1000/batch_size)-1
workers = 1 # Using GPU.
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
# Define the Keras model and compile it for training
model = models.Model(inputs=inputs, outputs=output_layer)

model.compile(optimizer=keras.optimizers.Adam(learning_rate), loss='categorical_crossentropy')

# Data iterators for loading the training and validation data
train_iter = data_iterator.BatchIteratorSimple(batch_size=batch_size,
                                               data_folder=os.path.join('..', 'data', 'train'),
                                               image_shape=image_shape,
                                               shift_aug=True)

val_iter = data_iterator.BatchIteratorSimple(batch_size=batch_size,
                                             data_folder=os.path.join('..', 'data', 'validation'),
                                             image_shape=image_shape)

logger_cb = plotting_tools.LoggerPlotter()
callbacks = [logger_cb]

model.fit_generator(train_iter,
                    steps_per_epoch = steps_per_epoch, # the number of batches per epoch,
                    epochs = num_epochs, # the number of epochs to train for,
                    validation_data = val_iter, # validation iterator
                    validation_steps = validation_steps, # the number of batches to validate on
                    callbacks=callbacks,
                    workers = workers)

# Save your trained model weights
weight_file_name = 'model_weights-1'
model_tools.save_network(model, weight_file_name)
score_all()

Feature layer (?, 160, 160, 3)
Encoder 1 (?, 80, 80, 32)
Encoder 2 (?, 40, 40, 64)
Encoder 3 (?, 20, 20, 128)
1x1 Conv (?, 20, 20, 128)
Decoder 1 (?, 40, 40, 128)
Decoder 2 (?, 80, 80, 64)
Decoder 3 (?, 160, 160, 32)
Output_layer (?, 160, 160, 3)
Epoch 1/40
127/128 [============================>.] - ETA: 0s - loss: 0.1822

128/128 [==============================] - 56s - loss: 0.1812 - val_loss: 0.1129
Epoch 2/40
127/128 [============================>.] - ETA: 0s - loss: 0.0460

128/128 [==============================] - 55s - loss: 0.0460 - val_loss: 0.1144
Epoch 3/40
127/128 [============================>.] - ETA: 0s - loss: 0.0397

128/128 [==============================] - 55s - loss: 0.0401 - val_loss: 0.0871
Epoch 4/40
127/128 [============================>.] - ETA: 0s - loss: 0.0378

128/128 [==============================] - 54s - loss: 0.0377 - val_loss: 0.0719
Epoch 5/40
127/128 [============================>.] - ETA: 0s - loss: 0.0333

128/128 [==============================] - 54s - loss: 0.0336 - val_loss: 0.0717
Epoch 6/40
127/128 [============================>.] - ETA: 0s - loss: 0.0336

128/128 [==============================] - 55s - loss: 0.0336 - val_loss: 0.0464
Epoch 7/40
127/128 [============================>.] - ETA: 0s - loss: 0.0330

128/128 [==============================] - 55s - loss: 0.0329 - val_loss: 0.0421
Epoch 8/40
127/128 [============================>.] - ETA: 0s - loss: 0.0299

128/128 [==============================] - 55s - loss: 0.0299 - val_loss: 0.0394
Epoch 9/40
127/128 [============================>.] - ETA: 0s - loss: 0.0289

128/128 [==============================] - 55s - loss: 0.0289 - val_loss: 0.0379
Epoch 10/40
127/128 [============================>.] - ETA: 0s - loss: 0.0293

# Run One
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""

image_hw = 160
image_shape = (image_hw, image_hw, 3)
inputs = layers.Input(image_shape)
num_classes = 3

# Call fcn_model()
output_layer = fcn_model(inputs, num_classes)

learning_rate = 0.001
batch_size = 32
num_epochs = 40
steps_per_epoch = int(4131/batch_size)-1#200
validation_steps = int(1000/batch_size)-1
workers = 1 # Using GPU.
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
# Define the Keras model and compile it for training
model = models.Model(inputs=inputs, outputs=output_layer)

model.compile(optimizer=keras.optimizers.Adam(learning_rate), loss='categorical_crossentropy')

# Data iterators for loading the training and validation data
train_iter = data_iterator.BatchIteratorSimple(batch_size=batch_size,
                                               data_folder=os.path.join('..', 'data', 'train'),
                                               image_shape=image_shape,
                                               shift_aug=True)

val_iter = data_iterator.BatchIteratorSimple(batch_size=batch_size,
                                             data_folder=os.path.join('..', 'data', 'validation'),
                                             image_shape=image_shape)

logger_cb = plotting_tools.LoggerPlotter()
callbacks = [logger_cb]

model.fit_generator(train_iter,
                    steps_per_epoch = steps_per_epoch, # the number of batches per epoch,
                    epochs = num_epochs, # the number of epochs to train for,
                    validation_data = val_iter, # validation iterator
                    validation_steps = validation_steps, # the number of batches to validate on
                    callbacks=callbacks,
                    workers = workers)

# Save your trained model weights
weight_file_name = 'model_weights-lr001'
model_tools.save_network(model, weight_file_name)
score_all()

Feature layer (?, 160, 160, 3)
Encoder 1 (?, 80, 80, 32)
Encoder 2 (?, 40, 40, 64)
Encoder 3 (?, 20, 20, 128)
1x1 Conv (?, 20, 20, 128)
Decoder 1 (?, 40, 40, 128)
Decoder 2 (?, 80, 80, 64)
Decoder 3 (?, 160, 160, 32)
Output_layer (?, 160, 160, 3)
Epoch 1/40
127/128 [============================>.] - ETA: 0s - loss: 0.7046

128/128 [==============================] - 55s - loss: 0.7013 - val_loss: 0.2762
Epoch 2/40
127/128 [============================>.] - ETA: 0s - loss: 0.1478

128/128 [==============================] - 54s - loss: 0.1474 - val_loss: 0.1129
Epoch 3/40
127/128 [============================>.] - ETA: 0s - loss: 0.0677

128/128 [==============================] - 54s - loss: 0.0677 - val_loss: 0.0988
Epoch 4/40
127/128 [============================>.] - ETA: 0s - loss: 0.0505

128/128 [==============================] - 54s - loss: 0.0505 - val_loss: 0.0977
Epoch 5/40
127/128 [============================>.] - ETA: 0s - loss: 0.0438

128/128 [==============================] - 54s - loss: 0.0437 - val_loss: 0.0520
Epoch 6/40
127/128 [============================>.] - ETA: 0s - loss: 0.0393

128/128 [==============================] - 54s - loss: 0.0393 - val_loss: 0.0456
Epoch 7/40
127/128 [============================>.] - ETA: 0s - loss: 0.0354

128/128 [==============================] - 54s - loss: 0.0354 - val_loss: 0.0486
Epoch 8/40
127/128 [============================>.] - ETA: 0s - loss: 0.0352

128/128 [==============================] - 54s - loss: 0.0352 - val_loss: 0.0415
Epoch 9/40
127/128 [============================>.] - ETA: 0s - loss: 0.0321

128/128 [==============================] - 54s - loss: 0.0329 - val_loss: 0.0315
Epoch 10/40
127/128 [============================>.] - ETA: 0s - loss: 0.0325

# Run One
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""

image_hw = 160
image_shape = (image_hw, image_hw, 3)
inputs = layers.Input(image_shape)
num_classes = 3

# Call fcn_model()
output_layer = fcn_model(inputs, num_classes)

learning_rate = 0.001
batch_size = 32
num_epochs = 80
steps_per_epoch = int(4131/batch_size)-1#200
validation_steps = int(1000/batch_size)-1
workers = 1 # Using GPU.
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
# Define the Keras model and compile it for training
model = models.Model(inputs=inputs, outputs=output_layer)

model.compile(optimizer=keras.optimizers.Adam(learning_rate), loss='categorical_crossentropy')

# Data iterators for loading the training and validation data
train_iter = data_iterator.BatchIteratorSimple(batch_size=batch_size,
                                               data_folder=os.path.join('..', 'data', 'train'),
                                               image_shape=image_shape,
                                               shift_aug=True)

val_iter = data_iterator.BatchIteratorSimple(batch_size=batch_size,
                                             data_folder=os.path.join('..', 'data', 'validation'),
                                             image_shape=image_shape)

logger_cb = plotting_tools.LoggerPlotter()
callbacks = [logger_cb]

model.fit_generator(train_iter,
                    steps_per_epoch = steps_per_epoch, # the number of batches per epoch,
                    epochs = num_epochs, # the number of epochs to train for,
                    validation_data = val_iter, # validation iterator
                    validation_steps = validation_steps, # the number of batches to validate on
                    callbacks=callbacks,
                    workers = workers)

# Save your trained model weights
weight_file_name = 'model_weights-lr001ep80'
model_tools.save_network(model, weight_file_name)
score_all()

Feature layer (?, 160, 160, 3)
Encoder 1 (?, 80, 80, 32)
Encoder 2 (?, 40, 40, 64)
Encoder 3 (?, 20, 20, 128)
1x1 Conv (?, 20, 20, 128)
Decoder 1 (?, 40, 40, 128)
Decoder 2 (?, 80, 80, 64)
Decoder 3 (?, 160, 160, 32)
Output_layer (?, 160, 160, 3)
Epoch 1/80
127/128 [============================>.] - ETA: 0s - loss: 0.6876

128/128 [==============================] - 55s - loss: 0.6843 - val_loss: 0.2691
Epoch 2/80
127/128 [============================>.] - ETA: 0s - loss: 0.1407

128/128 [==============================] - 55s - loss: 0.1402 - val_loss: 0.1229
Epoch 3/80
127/128 [============================>.] - ETA: 0s - loss: 0.0651

128/128 [==============================] - 54s - loss: 0.0649 - val_loss: 0.1051
Epoch 4/80
127/128 [============================>.] - ETA: 0s - loss: 0.0495

128/128 [==============================] - 55s - loss: 0.0494 - val_loss: 0.0782
Epoch 5/80
127/128 [============================>.] - ETA: 0s - loss: 0.0411

128/128 [==============================] - 54s - loss: 0.0410 - val_loss: 0.0566
Epoch 6/80
127/128 [============================>.] - ETA: 0s - loss: 0.0365

128/128 [==============================] - 55s - loss: 0.0365 - val_loss: 0.0394
Epoch 7/80
127/128 [============================>.] - ETA: 0s - loss: 0.0347

128/128 [==============================] - 54s - loss: 0.0347 - val_loss: 0.0409
Epoch 8/80
127/128 [============================>.] - ETA: 0s - loss: 0.0330

128/128 [==============================] - 55s - loss: 0.0329 - val_loss: 0.0324
Epoch 9/80
127/128 [============================>.] - ETA: 0s - loss: 0.0310

128/128 [==============================] - 55s - loss: 0.0312 - val_loss: 0.0493
Epoch 10/80
127/128 [============================>.] - ETA: 0s - loss: 0.0314

# Run One
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""

image_hw = 160
image_shape = (image_hw, image_hw, 3)
inputs = layers.Input(image_shape)
num_classes = 3

# Call fcn_model()
output_layer = fcn_model(inputs, num_classes)

learning_rate = 0.001
batch_size = 32
num_epochs = 80
steps_per_epoch = int(4131/batch_size)-1#200
validation_steps = int(1000/batch_size)-1
workers = 1 # Using GPU.
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
# Define the Keras model and compile it for training
model = models.Model(inputs=inputs, outputs=output_layer)

model.compile(optimizer=keras.optimizers.Adam(learning_rate), loss='categorical_crossentropy')

# Data iterators for loading the training and validation data
train_iter = data_iterator.BatchIteratorSimple(batch_size=batch_size,
                                               data_folder=os.path.join('..', 'data', 'train'),
                                               image_shape=image_shape,
                                               shift_aug=True)

val_iter = data_iterator.BatchIteratorSimple(batch_size=batch_size,
                                             data_folder=os.path.join('..', 'data', 'validation'),
                                             image_shape=image_shape)

logger_cb = plotting_tools.LoggerPlotter()
callbacks = [logger_cb]

model.fit_generator(train_iter,
                    steps_per_epoch = steps_per_epoch, # the number of batches per epoch,
                    epochs = num_epochs, # the number of epochs to train for,
                    validation_data = val_iter, # validation iterator
                    validation_steps = validation_steps, # the number of batches to validate on
                    callbacks=callbacks,
                    workers = workers)

# Save your trained model weights
weight_file_name = 'model_weights-lr001ep80'
model_tools.save_network(model, weight_file_name)
score_all()

Feature layer (?, 160, 160, 3)
Encoder 1 (?, 80, 80, 32)
Encoder 2 (?, 40, 40, 64)
Encoder 3 (?, 20, 20, 128)
1x1 Conv (?, 20, 20, 128)
Decoder 1 (?, 40, 40, 128)
Decoder 2 (?, 80, 80, 64)
Decoder 3 (?, 160, 160, 32)
Output_layer (?, 160, 160, 3)
Epoch 1/80
127/128 [============================>.] - ETA: 0s - loss: 0.6880

128/128 [==============================] - 58s - loss: 0.6847 - val_loss: 0.3110
Epoch 2/80
127/128 [============================>.] - ETA: 0s - loss: 0.1400

128/128 [==============================] - 54s - loss: 0.1395 - val_loss: 0.1327
Epoch 3/80
127/128 [============================>.] - ETA: 0s - loss: 0.0657

128/128 [==============================] - 55s - loss: 0.0656 - val_loss: 0.0906
Epoch 4/80
127/128 [============================>.] - ETA: 0s - loss: 0.0493

128/128 [==============================] - 54s - loss: 0.0498 - val_loss: 0.0820
Epoch 5/80
127/128 [============================>.] - ETA: 0s - loss: 0.0418

128/128 [==============================] - 54s - loss: 0.0418 - val_loss: 0.0567
Epoch 6/80
127/128 [============================>.] - ETA: 0s - loss: 0.0389

128/128 [==============================] - 54s - loss: 0.0389 - val_loss: 0.0445
Epoch 7/80
127/128 [============================>.] - ETA: 0s - loss: 0.0356

128/128 [==============================] - 54s - loss: 0.0356 - val_loss: 0.0424
Epoch 8/80
127/128 [============================>.] - ETA: 0s - loss: 0.0335

128/128 [==============================] - 54s - loss: 0.0336 - val_loss: 0.0337
Epoch 9/80
127/128 [============================>.] - ETA: 0s - loss: 0.0317

128/128 [==============================] - 54s - loss: 0.0317 - val_loss: 0.0501
Epoch 10/80
127/128 [============================>.] - ETA: 0s - loss: 0.0299

weight_file_name = 'model_weights-lr001ep80'
model_tools.save_network(model, weight_file_name)
score_all()

number of validation samples intersection over the union evaulated on 542
average intersection over union for background is 0.9952818098013742
average intersection over union for other people is 0.3475127989407102
average intersection over union for the hero is 0.89822017467698
number true positives: 539, number false positives: 0, number false negatives: 0
number of validation samples intersection over the union evaulated on 270
average intersection over union for background is 0.984562128278374
average intersection over union for other people is 0.6868302120677434
average intersection over union for the hero is 0.0
number true positives: 0, number false positives: 52, number false negatives: 0
number of validation samples intersection over the union evaulated on 322
average intersection over union for background is 0.9959821475895548
average intersection over union for other people is 0.42966077806067465
average intersection over union for the hero is 0.18309341584095531
number true positives: 132, number false positives: 4, number false negatives: 169
0.7488839285714286
0.540656795259
0.404889184842

# Run 3
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""

image_hw = 160
image_shape = (image_hw, image_hw, 3)
inputs = layers.Input(image_shape)
num_classes = 3

# Call fcn_model()
output_layer = fcn_model(inputs, num_classes)

learning_rate = 0.001
batch_size = 32
num_epochs = 80
steps_per_epoch = int(4131/batch_size)-1#200
validation_steps = int(1000/batch_size)-1
workers = 1 # Using GPU.
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
# Define the Keras model and compile it for training
model = models.Model(inputs=inputs, outputs=output_layer)

model.compile(optimizer=keras.optimizers.Adam(learning_rate), loss='categorical_crossentropy')

# Data iterators for loading the training and validation data
train_iter = data_iterator.BatchIteratorSimple(batch_size=batch_size,
                                               data_folder=os.path.join('..', 'data', 'train'),
                                               image_shape=image_shape,
                                               shift_aug=True)

val_iter = data_iterator.BatchIteratorSimple(batch_size=batch_size,
                                             data_folder=os.path.join('..', 'data', 'validation'),
                                             image_shape=image_shape)

logger_cb = plotting_tools.LoggerPlotter()
callbacks = [logger_cb]

model.fit_generator(train_iter,
                    steps_per_epoch = steps_per_epoch, # the number of batches per epoch,
                    epochs = num_epochs, # the number of epochs to train for,
                    validation_data = val_iter, # validation iterator
                    validation_steps = validation_steps, # the number of batches to validate on
                    callbacks=callbacks,
                    workers = workers)

# Save your trained model weights
weight_file_name = 'model_weights-lr001ep80_2'
model_tools.save_network(model, weight_file_name)
score_all()

Feature layer (?, 160, 160, 3)
Encoder 1 (?, 80, 80, 32)
Encoder 2 (?, 40, 40, 64)
Encoder 3 (?, 20, 20, 128)
1x1 Conv (?, 20, 20, 128)
Decoder 1 (?, 40, 40, 128)
Decoder 2 (?, 80, 80, 64)
Decoder 3 (?, 160, 160, 32)
Output_layer (?, 160, 160, 3)
Epoch 1/80
127/128 [============================>.] - ETA: 0s - loss: 0.7026

128/128 [==============================] - 55s - loss: 0.6994 - val_loss: 0.2932
Epoch 2/80
127/128 [============================>.] - ETA: 0s - loss: 0.1472

128/128 [==============================] - 54s - loss: 0.1470 - val_loss: 0.1135
Epoch 3/80
127/128 [============================>.] - ETA: 0s - loss: 0.0740

128/128 [==============================] - 54s - loss: 0.0739 - val_loss: 0.1053
Epoch 4/80
127/128 [============================>.] - ETA: 0s - loss: 0.0531

128/128 [==============================] - 54s - loss: 0.0530 - val_loss: 0.0803
Epoch 5/80
127/128 [============================>.] - ETA: 0s - loss: 0.0437

128/128 [==============================] - 54s - loss: 0.0438 - val_loss: 0.0691
Epoch 6/80
127/128 [============================>.] - ETA: 0s - loss: 0.0399

128/128 [==============================] - 54s - loss: 0.0399 - val_loss: 0.0358
Epoch 7/80
127/128 [============================>.] - ETA: 0s - loss: 0.0362

128/128 [==============================] - 54s - loss: 0.0362 - val_loss: 0.0432
Epoch 8/80
127/128 [============================>.] - ETA: 0s - loss: 0.0332

128/128 [==============================] - 54s - loss: 0.0332 - val_loss: 0.0391
Epoch 9/80
127/128 [============================>.] - ETA: 0s - loss: 0.0329

128/128 [==============================] - 54s - loss: 0.0330 - val_loss: 0.0438
Epoch 10/80
127/128 [============================>.] - ETA: 0s - loss: 0.0308

"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""

image_hw = 160
image_shape = (image_hw, image_hw, 3)
inputs = layers.Input(image_shape)
num_classes = 3

# Call fcn_model()
output_layer = fcn_model(inputs, num_classes)

Feature layer (?, 160, 160, 3)
Encoder 1 (?, 80, 80, 32)
Encoder 2 (?, 40, 40, 64)
Encoder 3 (?, 20, 20, 128)
1x1 Conv (?, 20, 20, 128)
Decoder 1 (?, 40, 40, 128)
Decoder 2 (?, 80, 80, 64)
Decoder 3 (?, 160, 160, 32)
Output_layer (?, 160, 160, 3)

Hyperparameters¶

Define and tune your hyperparameters.

batch_size: number of training samples/images that get propagated through the network in a single pass.
num_epochs: number of times the entire training dataset gets propagated through the network.
steps_per_epoch: number of batches of training images that go through the network in 1 epoch. We have provided you with a default value. One recommended value to try would be based on the total number of images in training dataset divided by the batch_size.
validation_steps: number of batches of validation images that go through the network in 1 epoch. This is similar to steps_per_epoch, except validation_steps is for the validation dataset. We have provided you with a default value for this as well.
workers: maximum number of processes to spin up. This can affect your training speed and is dependent on your hardware. We have provided a recommended value to work with.

learning_rate = 0.001
batch_size = 32
num_epochs = 80
steps_per_epoch = int(4131/batch_size)-1
validation_steps = int(1000/batch_size)-1
workers = 2

"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""

from workspace_utils import active_session
# Keeping Your Session Active
with active_session():
    # Define the Keras model and compile it for training
    model = models.Model(inputs=inputs, outputs=output_layer)

    model.compile(optimizer=keras.optimizers.Adam(learning_rate), loss='categorical_crossentropy')

    # Data iterators for loading the training and validation data
    train_iter = data_iterator.BatchIteratorSimple(batch_size=batch_size,
                                                   data_folder=os.path.join('..', 'data', 'train'),
                                                   image_shape=image_shape,
                                                   shift_aug=True)

    val_iter = data_iterator.BatchIteratorSimple(batch_size=batch_size,
                                                 data_folder=os.path.join('..', 'data', 'validation'),
                                                 image_shape=image_shape)

    logger_cb = plotting_tools.LoggerPlotter()
    callbacks = [logger_cb]

    model.fit_generator(train_iter,
                        steps_per_epoch = steps_per_epoch, # the number of batches per epoch,
                        epochs = num_epochs, # the number of epochs to train for,
                        validation_data = val_iter, # validation iterator
                        validation_steps = validation_steps, # the number of batches to validate on
                        callbacks=callbacks,
                        workers = workers)

Epoch 1/80
199/200 [============================>.] - ETA: 0s - loss: 0.0179

200/200 [==============================] - 82s - loss: 0.0179 - val_loss: 0.0335
Epoch 2/80
199/200 [============================>.] - ETA: 0s - loss: 0.0183

200/200 [==============================] - 81s - loss: 0.0183 - val_loss: 0.0298
Epoch 3/80
199/200 [============================>.] - ETA: 0s - loss: 0.0169

200/200 [==============================] - 81s - loss: 0.0169 - val_loss: 0.0269
Epoch 4/80
199/200 [============================>.] - ETA: 0s - loss: 0.0177

200/200 [==============================] - 81s - loss: 0.0177 - val_loss: 0.0247
Epoch 5/80
199/200 [============================>.] - ETA: 0s - loss: 0.0163

200/200 [==============================] - 81s - loss: 0.0163 - val_loss: 0.0332
Epoch 6/80
199/200 [============================>.] - ETA: 0s - loss: 0.0164

200/200 [==============================] - 81s - loss: 0.0164 - val_loss: 0.0320
Epoch 7/80
199/200 [============================>.] - ETA: 0s - loss: 0.0161

200/200 [==============================] - 81s - loss: 0.0161 - val_loss: 0.0202
Epoch 8/80
199/200 [============================>.] - ETA: 0s - loss: 0.0152

200/200 [==============================] - 81s - loss: 0.0153 - val_loss: 0.0273
Epoch 9/80
 70/200 [=========>....................] - ETA: 47s - loss: 0.0167

# Save your trained model weights
weight_file_name = 'model_weights_1_29_1_49t_lr01_batch_size_100_epochs_40_1sep_in_dec'
model_tools.save_network(model, weight_file_name)

Prediction ¶

Now that you have your model trained and saved, you can make predictions on your validation dataset. These predictions can be compared to the mask images, which are the ground truth labels, to evaluate how well your model is doing under different conditions.

There are three different predictions available from the helper code provided:

patrol_with_targ: Test how well the network can detect the hero from a distance.
patrol_non_targ: Test how often the network makes a mistake and identifies the wrong person as the target.
following_images: Test how well the network can identify the target while following them.

# If you need to load a model which you previously trained you can uncomment the codeline that calls the function below.

# weight_file_name = 'model_weights'
# restored_model = model_tools.load_network(weight_file_name)

The following cell will write predictions to files and return paths to the appropriate directories. The run_num parameter is used to define or group all the data for a particular model run. You can change it for different runs. For example, 'run_1', 'run_2' etc.

run_num = 'run_1'

val_with_targ, pred_with_targ = model_tools.write_predictions_grade_set(model,
                                        run_num,'patrol_with_targ', 'sample_evaluation_data') 

val_no_targ, pred_no_targ = model_tools.write_predictions_grade_set(model, 
                                        run_num,'patrol_non_targ', 'sample_evaluation_data') 

val_following, pred_following = model_tools.write_predictions_grade_set(model,
                                        run_num,'following_images', 'sample_evaluation_data')

Now lets look at your predictions, and compare them to the ground truth labels and original images. Run each of the following cells to visualize some sample images from the predictions in the validation set.

# images while following the target
im_files = plotting_tools.get_im_file_sample('sample_evaluation_data','following_images', run_num) 
for i in range(3):
    im_tuple = plotting_tools.load_images(im_files[i])
    plotting_tools.show_images(im_tuple)

# images while at patrol without target
im_files = plotting_tools.get_im_file_sample('sample_evaluation_data','patrol_non_targ', run_num) 
for i in range(3):
    im_tuple = plotting_tools.load_images(im_files[i])
    plotting_tools.show_images(im_tuple)

   
# images while at patrol with target
im_files = plotting_tools.get_im_file_sample('sample_evaluation_data','patrol_with_targ', run_num) 
for i in range(3):
    im_tuple = plotting_tools.load_images(im_files[i])
    plotting_tools.show_images(im_tuple)

Evaluation ¶

Evaluate your model! The following cells include several different scores to help you evaluate your model under the different conditions discussed during the Prediction step.

# Scores for while the quad is following behind the target. 
true_pos1, false_pos1, false_neg1, iou1 = scoring_utils.score_run_iou(val_following, pred_following)

number of validation samples intersection over the union evaulated on 542
average intersection over union for background is 0.9954890433886608
average intersection over union for other people is 0.3458723724673306
average intersection over union for the hero is 0.9058804435865931
number true positives: 539, number false positives: 1, number false negatives: 0

# Scores for images while the quad is on patrol and the target is not visable
true_pos2, false_pos2, false_neg2, iou2 = scoring_utils.score_run_iou(val_no_targ, pred_no_targ)

number of validation samples intersection over the union evaulated on 270
average intersection over union for background is 0.9882562295939652
average intersection over union for other people is 0.7501523127965406
average intersection over union for the hero is 0.0
number true positives: 0, number false positives: 75, number false negatives: 0

# This score measures how well the neural network can detect the target from far away
true_pos3, false_pos3, false_neg3, iou3 = scoring_utils.score_run_iou(val_with_targ, pred_with_targ)

number of validation samples intersection over the union evaulated on 322
average intersection over union for background is 0.9965715649439747
average intersection over union for other people is 0.4425629432887988
average intersection over union for the hero is 0.23930189225590606
number true positives: 154, number false positives: 4, number false negatives: 147

# Sum all the true positives, etc from the three datasets to get a weight for the score
true_pos = true_pos1 + true_pos2 + true_pos3
false_pos = false_pos1 + false_pos2 + false_pos3
false_neg = false_neg1 + false_neg2 + false_neg3

weight = true_pos/(true_pos+false_neg+false_pos)
print(weight)

0.7532608695652174

# The IoU for the dataset that never includes the hero is excluded from grading
final_IoU = (iou1 + iou3)/2
print(final_IoU)

0.572591167921

# And the final grade score is 
final_score = final_IoU * weight
print(final_score)

0.431310521054